Guest post by Mike Jonas,

Maybe, after all the attention being paid to the Wuhan virus, it’s time to do a bit of climate science again.

I have submitted a paper to a peer-reviewed journal, and, remarkably, the journal says it is happy for me to put it up on the web while it is in review, so here it is. But it hasn’t been easy getting to this point: I think that what I describe in the paper is of great significance to climate science, as you can tell from some of the wording in the paper, so I was keen to get it published in a peer-reviewed journal – the IPCC are supposed to work only from peer-reviewed papers. The first journal I submitted it to (in November 2019) took two months to tell me that the paper was outside the scope of the journal – a curious claim, I thought, since one of the relevant papers that I was citing had been published in that journal. The second journal took four months to tell me that it hadn’t yet been seen by a reviewer and it needed a different format for references, plus a few other minor format changes. So I withdrew the paper. I have made modest improvements to the paper over that period, but I doubt they will make much difference to reviews.  Anyway, here’s hoping for third time lucky!

Here is the paper (and after the paper I’ve made some comments about how it relates to Richard Lindzen’s “Iris” theory):

– – – – – the paper: – – – – –

Cloud Feedback, if there is any, is Negative

Author: M Jonas

Affiliations: None

ABSTRACT

Virtually all the climate models referenced by the IPCC show a strong positive cloud feedback. Cloud feedback is the process by which a changing surface temperature affects cloud cover, which in turn affects surface temperature. In this paper, all monthly satellite data for sea surface temperatures and cloud cover over the oceans, for the whole available period of July 1986 to June 2017, is analysed, in order to test this feature of the climate models. As expected, the trends for the overall period are of rising sea surface temperatures and of falling cloud cover. But the analysis also shows an unexpected relationship between sea surface temperature and cloud cover: increases in sea surface temperature are associated with increases – not decreases – in cloud cover over the next few months. Moreover, the cloud cover increases tend to intercept a greater proportion of incoming solar radiation than they do of outgoing ocean radiation. The inevitable conclusion is that cloud feedback is negative. In any case, the observed reduction in cloud cover over the oceans between 1986 and 2017 could not have been a feedback from rising temperature. The implications for climate models are devastating.

KEYWORDS: Climate, Clouds, Cloud Feedback, SST, Sea Surface Temperature, Ocean

ABBREVIATIONS

dCloud – year-on-year change in cloud cover, dSST – year-on-year change in SST, GISS – (NASA) Goddard Institute for Space Studies, IPCC – Intergovernmental Panel on Climate Change, ISCCP – International Satellite Cloud Climatology Project, NASA – (US) National Aeronautics and Space Administration, NOAA – (US) National Oceanic and Atmospheric Administration, SST – Sea Surface Temperature.

1. INTRODUCTION

The ocean covers 70% of the global surface and stores more heat in the uppermost 3 metres than the entire atmosphere, so the key to understanding global climate change is inextricably linked to the ocean [7].

Spencer and Braswell [13], using satellite data and models, reported that atmospheric feedback diagnosis of the climate system remains an unsolved problem. They state: “The magnitude of the surface temperature response of the climate system to an imposed radiative energy imbalance remains just as uncertain today as it was decades ago“.  By using satellite data only, over a longer period, this study resolves one of the major feedback issues, namely the sign of cloud feedback. The study uses empirical data only. In particular, it uses all of the period with data, it uses the whole global ocean area, and it does not use any models.

The IPCC [5] defines Climate Feedback as an interaction between processes when “the result of an initial process triggers changes in a second process that in turn influences the initial one“, and states that a way to quantify it is as “the response of the climate system to a global surface temperature change“.  Cloud feedback is thus the process by which a changing surface temperature affects cloud cover, which in turn affects surface temperature. Note that cloud feedback does not depend on the original cause of the changing surface temperature.

Feedback response to temperature has been difficult to separate from other effects on temperature as the initial process is “nearly simultaneous with the temperature change“, but for the second process “there is a substantial time lag between forcing and the temperature response due to the heat capacity of the ocean” [13].

This study’s aim is to see if the effect of sea surface temperature (SST) on cloud cover can be detected. It does this by comparing changes in SST with later changes in cloud cover.

2. METHOD

Gridded monthly SST data (deg C) was downloaded from NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, in the USA National Oceanic and Atmospheric Administration (NOAA) [10]. Each grid cell is 1 degree latitude by 1 degree longitude.

Equal-area monthly cloud data (cloud cover % and IR optical depth) was downloaded from the International Satellite Cloud Climatology Project (ISCCP) in the NASA Goddard Institute for Space Studies (GISS), New York, NY [11]. Only data over the ocean was used. ie,  land and coastal data was not used.

Data from the two datasets was used only for the months and areas for which there was both cloud data and complete SST data. SST data was missing in locations and months where there was some sea ice.

For each month, SST and cloud cover were averaged over the total ocean area. For SST, individual temperature readings were converted into Kelvin and raised to the 4th power before averaging, then converted back to degrees C. This ensured that the calculated average SSTs related correctly to outward radiation from the sea surface.

For each month after the first year, the global year-on-year change in SST was calculated (referred to here as “dSST”). These were then averaged over various numbers of months, and each average was recorded against the end month of the period being averaged.  The global year-on-year change in cloud cover (“dCloud”) was averaged for the same number of months but for later periods that did not overlap the period used for dSST. For comparison purposes, dCloud was recorded for the same month as the related dSST, for each combination of number of months averaged and number of months later. The linear trend of dCloud against dSST was then calculated for each combination using the standard spreadsheet Trend function (NB. the trend was of dCloud against dSST, not against time).

3. RATIONALE

As stated by Spencer and Braswell [13],  feedback response to temperature is “nearly simultaneous with the temperature change“. A change in SST should therefore result in a change in cloud cover (ie, cloud feedback) very soon after. There is therefore a possibility that cloud feedback can be detected in the monthly data. The relevant comparison to make is between SST change and cloud cover change in the following month or months.

This study uses year-on-year changes, and treats all calendar months equally, in order to avoid any possibility of seasonal effects. It avoids using overlapping periods for dSST and dCloud, in order to minimise any possibility of clouds’ influence on SST affecting results. And because a rapid effect is being looked for, it limits each overall period being looked at (from first dSST month to last dCloud month) to no more than 12 months.

The effect of cloud cover change is unlikely to show up in SST data within a month or within a few months [13], so comparing data the other way round – cloud cover change against SST change in the following month or months – is unlikely to be effective. However, clouds are known have a cooling effect on climate, see for example Pokrovsky [8] which calculates a (global) 0.07C warming effect for each 1% decrease in cloud cover.

4. RESULTS

4.1 Cloud Cover

In all cases with dCloud up to 9 months later than dSST, the linear trend of dCloud against dSST was positive, and this was a strong and consistent pattern. In other words, the months with higher dSST tended to be followed by months with higher dCloud, and lower dSST tended to be followed by lower dCloud.

The linear trends of dCloud against dSST are shown in Figure 1, and are shown graphically in Figure 2.

Linear trends of global average dCloud vs dSST – unweighted (ie, by equal area) +/- 2*Sigma

Months later:	1 mth	2 mths	3 mths	4 mths	5 mths	6 mths	7 mths	8 mths	9 mths	10 mths	11 mths
Months averaged
1 mth	1.21 +/- 0.68	1.46 +/- 0.67	1.74 +/- 0.67	1.34 +/- 0.68	1.23 +/- 0.68	1.25 +/- 0.68	0.91 +/- 0.69	0.95 +/- 0.69	0.91 +/- 0.70	0.63 +/- 0.70	0.07 +/- 0.71
2 mths		1.62 +/- 0.58	1.72 +/- 0.58	1.56 +/- 0.59	1.41 +/- 0.59	1.31 +/- 0.59	1.14 +/- 0.60	1.05 +/- 0.60	0.96 +/- 0.61	0.63 +/- 0.61
3 mths			1.74 +/- 0.55	1.67 +/- 0.55	1.55 +/- 0.56	1.41 +/- 0.56	1.27 +/- 0.56	1.15 +/- 0.57	0.93 +/- 0.58
4 mths				1.73 +/- 0.53	1.65 +/- 0.53	1.53 +/- 0.54	1.36 +/- 0.54	1.18 +/- 0.55
5 mths					1.72 +/- 0.52	1.61 +/- 0.53	1.43 +/- 0.53
6 mths						1.65 +/- 0.52

Figure 1. Linear trends of global average dCloud against dSST, with 95% confidence levels. Each is averaged over the number of months given under “Months averaged”. The dCloud values are for the given number of months later.

Figure 2. As Figure 1, graphically. The outer range of the 95% confidence levels is shown.

Note that the trends can be expected to drop off as ‘Months Later’ increases, because of influence from the intervening months.

The trends are clearly visible in charts of dCloud vs dSST, and reflect the body of data not just some outliers. Some examples are shown in Figures 3, 4, 5.

Figure 3. dSST for one month, vs dCloud the following month.

Figure 4. dSST averaged over 3 months, vs dCloud averaged over the following 3 months.

Figure 5. dSST averaged over 6 months, vs dCloud averaged over the following 6 months.

The charts in Figures 3, 4, 5 are all to the same scale.

4.2 Cloud Opacity

Cloud with greater optical depth intercepts more radiation – both incoming solar radiation and outgoing ocean radiation. The above analysis was re-calculated with clouds weighted by opacity (opaqueness). Opacity, as used in this study, is derived from optical depth as follows:

Optical Depth d is given by

d = ln(Fr/Ft)

where

d is optical depth,

Fr is flux received,

Ft = flux transmitted.

The proportion q of radiation intercepted by cloud –

q = (Fr–Ft)/Fr

– is referred to here as “opacity”. q can be derived from the formula for d as

q = 1 – e^(-d)

Opacity q can legitimately be arithmetically averaged across sets of clouds.

When cloud data is weighted by opacity, the table of dCloud against dSST becomes:

Linear trends of global average dCloud vs dSST – weighted by cloud opacity +/- 2*Sigma

Months later:	1 mth	2 mths	3 mths	4 mths	5 mths	6 mths	7 mths	8 mths	9 mths	10 mths	11 mths
Months averaged
1 mth	1.44 +/- 0.64	1.51 +/- 0.63	1.85 +/- 0.63	1.45 +/- 0.64	1.26 +/- 0.64	1.29 +/- 0.64	0.86 +/- 0.65	0.90 +/- 0.65	0.88 +/- 0.66	0.63 +/- 0.66	-0.02 +/- 0.67
2 mths		1.75 +/- 0.53	1.84 +/- 0.53	1.66 +/- 0.54	1.47 +/- 0.54	1.32 +/- 0.55	1.11 +/- 0.55	1.00 +/- 0.56	0.93 +/- 0.56	0.60 +/- 0.57
3 mths			1.87 +/- 0.49	1.77 +/- 0.50	1.61 +/- 0.50	1.43 +/- 0.51	1.25 +/- 0.51	1.11 +/- 0.52	0.89 +/- 0.52
4 mths				1.83 +/- 0.47	1.70 +/- 0.48	1.55 +/- 0.48	1.35 +/- 0.49	1.15 +/- 0.50
5 mths					1.78 +/- 0.46	1.64 +/- 0.46	1.42 +/- 0.47
6 mths						1.68 +/- 0.46

Figure 6. Linear trends of global average dCloud against dSST, with 95% confidence levels, as in Figure 1 but with cloud cover weighted by opacity.

As can be seen from Figures 1 and 6, the increased cloud cover associated with increased SST tends also to have greater opacity for several months. There is therefore not just an increased area of cloud, there is also an increased amount of cloud (as judged by its ability to intercept radiation).

4.3 Effect on incoming and outgoing radiation

In order to determine the relative effect of the cloud changes on incoming and outgoing radiation, the analysis was re-calculated with clouds weighted (a) by opacity multiplied by incoming solar radiation (monthly solar radiation data from [2]), and (b) by opacity multiplied by outgoing ocean radiation (from SST). Note that a difference between (a) and (b) would come primarily from the different distributions of incoming solar radiation and outgoing ocean radiation by latitude, not from variations in solar output.

The table of dCloud against dSST then becomes as shown in Figures 7 and 8.

Linear trends of global average dCloud vs dSST – weighted by cloud opacity and solar radiation +/- 2*Sigma

Months later:	1 mth	2 mths	3 mths	4 mths	5 mths	6 mths	7 mths	8 mths	9 mths	10 mths	11 mths
Months averaged
1 mth	1.60 +/- 0.73	1.79 +/- 0.72	2.25 +/- 0.71	1.91 +/- 0.72	1.68 +/- 0.73	1.76 +/- 0.73	1.16 +/- 0.74	1.22 +/- 0.74	1.22 +/- 0.75	0.99 +/- 0.75	0.16 +/- 0.76
2 mths		2.07 +/- 0.60	2.27 +/- 0.59	2.15 +/- 0.60	1.97 +/- 0.61	1.79 +/- 0.61	1.51 +/- 0.62	1.37 +/- 0.63	1.31 +/- 0.63	0.95 +/- 0.64
3 mths			2.31 +/- 0.55	2.28 +/- 0.55	2.13 +/- 0.56	1.92 +/- 0.57	1.70 +/- 0.58	1.54 +/- 0.58	1.28 +/- 0.59
4 mths				2.34 +/- 0.53	2.23 +/- 0.53	2.07 +/- 0.54	1.84 +/- 0.55	1.60 +/- 0.56
5 mths					2.31 +/- 0.51	2.17 +/- 0.51	1.93 +/- 0.53
6 mths						2.22 +/- 0.50

Figure 7. Linear trends of global average dCloud against dSST, with 95% confidence levels, as in Figure 1 but with cloud cover weighted by opacity multiplied by incoming solar radiation.

Linear trends of global average dCloud vs dSST – weighted by cloud opacity and ocean radiation +/- 2*Sigma

Months later:	1 mth	2 mths	3 mths	4 mths	5 mths	6 mths	7 mths	8 mths	9 mths	10 mths	11 mths
Months averaged
1 mth	1.44 +/- 0.68	1.56 +/- 0.67	1.91 +/- 0.67	1.52 +/- 0.68	1.33 +/- 0.68	1.38 +/- 0.68	0.93 +/- 0.69	0.99 +/- 0.69	0.96 +/- 0.70	0.71 +/- 0.70	0.02 +/- 0.71
2 mths		1.80 +/- 0.56	1.91 +/- 0.56	1.74 +/- 0.57	1.56 +/- 0.57	1.42 +/- 0.58	1.21 +/- 0.58	1.10 +/- 0.59	1.03 +/- 0.59	0.68 +/- 0.60
3 mths			1.94 +/- 0.52	1.86 +/- 0.53	1.71 +/- 0.53	1.53 +/- 0.54	1.36 +/- 0.54	1.22 +/- 0.55	0.99 +/- 0.55
4 mths				1.92 +/- 0.50	1.81 +/- 0.50	1.66 +/- 0.51	1.47 +/- 0.51	1.26 +/- 0.52
5 mths					1.89 +/- 0.48	1.75 +/- 0.49	1.54 +/- 0.50
6 mths						1.80 +/- 0.48

Figure 8. Linear trends of global average dCloud against dSST, with 95% confidence levels, as in Figure 1 but with cloud cover weighted by opacity multiplied by outgoing ocean radiation.

For the first few months, all the trends weighted by incoming solar radiation are higher than those weighted by outgoing ocean radiation. This shows that the increases in cloud cover associated with increases in SST have a larger effect on incoming radiation than on outgoing radiation. In other words, they are net cooling.

5. CONCLUSION

The data shows that there is a positive correlation between changes in SST and later changes in cloud cover. As stated above, it is known that clouds have a global cooling effect, and the analysis has shown that as the changes in cloud cover associated with changes in SST increase, they do indeed have a larger cooling effect. The inescapable conclusion is that any cloud feedback is negative.

This conclusion has profound implications for climate models. All of the models referenced by the IPCC are parameterised with positive cloud feedback – “the GCMs all predict a positive cloud feedback” – and they attribute nearly half of all anthropogenic global warming to cloud feedback [9]. This is a possible reason for most climate models overestimating global warming [4] [12], although Anagnostopoulos [1] says that there are much more significant problems.

A further conclusion with profound implications for the climate models is that the observed reduction in cloud cover between 1986 and 2017 was not a feedback from rising temperatures and that it was powerful enough to eventually override any negative cloud feedback. The IPCC report [9] suggests that the models do not recognise the possibility that clouds can have any behaviour other than as a feedback:- in Key Uncertainties they only say of clouds: “Large uncertainties remain about how clouds might respond to global climate change“. This study shows that clouds must have important behaviour that is not included in the models.

The implication for the climate models is devastating. It must be questioned whether they are fit for purpose.

6. DISCUSSION

6.1 Interpretation

The most reasonable interpretations would appear to be:

1. The changes in cloud cover as in Figure 1 are caused by the changes in SST. If this is the case, then cloud feedback is negative.
   or
2. The changes in cloud cover and the changes in SST are both caused by some unknown factor (ie, correlation is not causation). If this is the case, then cloud feedback is zero.
   or
3. A combination of 1 and 2 applies. If this is the case, then cloud feedback is not as negative as in 1 but it is still negative.

6.2 Other Interpretations

Any interpretation other than the above seems extraordinarily unlikely, and in any case would completely invalidate the climate models referenced by the IPCC. For example:

(a) If an increase in SST actually caused there to be less clouds, then some unknown factor or factors would have to be operating in the opposite direction with greater effect. ie, some unknown factor or factors would have to be creating more clouds after an increase in SST even though an increase in SST was causing there to be less clouds.

(b) If an increase in SST does cause more clouds, but if in fact clouds have a warming effect, not cooling, then some unknown factor or factors would have to be operating over the longer term to both reduce cloud cover and to increase SSTs.

As stated above, both these possibilities are extremely unlikely, and can surely be discounted. Alternative (b) in particular appears to be ruled out anyway by the above analysis using incoming and outgoing radiation weightings.

6.3 Quantification

It would be tempting to quantify the negative cloud feedback using the linear trends between dCloud and dSST as reported above and the linear trends of SST and cloud cover over the period that was analysed.

Such quantification would be unjustifiable at this stage, because until the mechanisms involved are reasonably well understood, it would be far too unreliable. There is an obvious possible mechanism for cloud feedback being negative:- as oceans warm they release more water vapour into the atmosphere, which then forms more clouds over the next few months. This concept is supported by the observation by Durack [3] of an increased hydrological cycle. However, the mechanism needs to be evaluated against the findings here before it can be assumed to be the relevant mechanism.

6.4 What Next?

Quantification of cloud feedback is clearly needed, as in 6.3 Quantification.

But further investigation into the behaviour of clouds is also needed:

This study shows that the observed reduction in cloud cover between 1986 and 2017 was not a feedback from rising temperatures. It follows that some of the increase in SST over this period could have been caused by independent cloud cover reduction. Finding out what caused the cloud cover reduction would be a major step forward for climate science. Some work has already been done, eg. [6] [14].

On a more general level:- It is abundantly clear that the climate models cannot represent the behaviour of clouds. It must therefore be seriously questioned whether the models, as currently structured, will ever be of any value at all for predicting future climate. It would be reasonable to consider the advice given by Anagnostopoulos [1] that a paradigm shift is needed.

6.5 Overall Trends

Charts of SST (Deg C) and Cloud cover (%) are given in Figures 9 and 10. These support the statement in the Abstract: “As expected, the trends for the overall period are of rising sea surface temperatures and of falling cloud cover.”.

Figure 9. SST data over the study period, with linear trend.

Figure 10. Cloud cover data over the study period, with linear trend.

The sources of the SST and cloud data are given in 2. METHOD.

Acknowledgements

Most of the data was accessed using Panoply software developed and provided free of charge by NASA Goddard Institute for Space Studies.

References

1. Anagnostopoulos GG et al 2010: A comparison of local and aggregated climate model outputs with observed data. Hydrological Sciences Journal 55(7), 1094–1110.

https://doi.org/10.1080/02626667.2010.513518

2. Coddington et al 2015: NOAA Climate Data Record (CDR) of Total Solar Irradiance (TSI), NRLTSI Version 2 [monthly TSI]. NOAA National Centers for Environmental Information doi:10.7289/V55B00C1. Data downloaded Jan 2020 from https://ift.tt/2XCkW5j

3. Durack et al 2012: Ocean Salinities Reveal Strong Global Water Cycle Intensification During 1950 to 2000. Science 27 Apr 2012 Vol. 336, Issue 6080, pp. 455-458 DOI: 10.1126/science.1212222

4. Fyfe J et al 2013: Overestimated global warming over the past 20 years. Nature Climate Change 3,767–769 (2013) doi:10.1038/nclimate1972

5. IPCC 2007: Annex I, Glossary, in Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.)]. IPCC, Geneva, Switzerland, 104 pp.

6. Kamide Y 2007: Effects of the Solar Cycle on the Earth’s Atmosphere (in Handbook of the Solar-Terrestrial Environment). Springer, Berlin, Heidelberg.

https://doi.org/10.1007/978-3-540-46315-3_18

7. Nagaraja MP 2019: Climate Variability. NASA Science. Accessed 18 Nov 2019 at https://ift.tt/377EVfh

8. Pokrovsky OM 2019: Cloud Changes in the Period of Global Warming: the Results of the International Satellite Project. Russian Academy of Sciences;

https://doi.org/10.31857/S0205-9614201913-13

9. Randall DA et al 2007: [Climate] Models and Their Evaluation. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon S et al (editors.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

10. Reynolds RW et al 2002: An improved in situ and satellite SST analysis for climate. Journal of Climate, 15, 1609-1625.  Data accessed 26-30 Aug 2019 at https://ift.tt/2AzDHeY

11. Rossow WB and Schiffer RA 1999: Advances in understanding clouds from ISCCP. Bulletin of the American Meteorological Society, 80, 2261-2288, doi:10.1175/1520-0477(1999)080<2261:AIUCFI>2.0.CO;2. Data accessed 26 Aug 2019 at https://ift.tt/2MBGDMx

12. Spencer RW 2013: STILL Epic Fail: 73 Climate Models vs. Measurements, Running 5-Year Means.

13. Spencer RW and Braswell WD 2011: On the Misdiagnosis of Surface Temperature Feedbacks from Variations in Earth’s Radiant Energy Balance. Remote Sensing 2011, 3(8), 1603-1613; https://ift.tt/2Y7Hwl7

14. Svensmark J et al 2016: The response of clouds and aerosols to cosmic ray decreases, Journal of Geophysical Research – Space Physics, 2016, DOI: 10.1002/2016JA022689.

– – – – – end of paper – – – – –

The paper may have to be changed to satisfy the review process. But in the meantime, the paper presented above is the exact paper that has been presented to the journal.

“Iris” Theory

Richard Lindzen et al hypothesised in September 2000 that Earth might have an “Adaptive Infrared Iris” which provides a significant negative feedback to surface temperature changes – the “Iris” theory. If I have understood that paper correctly, then my paper above does not support the “Iris” theory, because the “Iris” theory is based on surface warming resulting in less clouds of a particular type at the tropics, whereas I found that the data associated cloud increase with surface temperature increase.

NB, my findings don’t disprove the “Iris” theory either, because the “Iris” theory is based on specific cloud reduction in a limited area, whereas my analysis uses only global SST and cloud data.

In January 2002, Bing Lin et al argued against the “Iris” theory, saying that the cloud changes were actually warming not cooling, and therefore that the feedback referred to was positive not negative.

My paper provides no more support for Bing Lin’s argument than it does for Richard Lindzen’s, for the same reason: Bing Lin refers to the same cloud reduction.

Incidentally, arguments based on different types of cloud won’t work if anyone tries to use them to disprove my findings. That’s because surface temperature increases are followed by cloud increases. If the cloud type involved somehow manages to make that a positive feedback – in spite of my finding re cloud opacity – then the positive feedback will create even more cloud which will deliver even more positive feedback, etc, etc. But … in the longer term, as surface temperature increases, there’s less cloud, not more, so my argument in 6.2 (b) applies.

And one final comment: All I have done is to analyse the data. I don’t provide any theories, I don’t use any models, and I don’t use any sophisticated statistical tricks or cherry-picking to manipulate the data into a dubious finding. All the available data is used, and the only process used is simple weighted (and unweighted) averaging. The chart of cloud change against temperature change goes the “wrong” way. Period.